Timeline for Number of parameters in an artificial neural network for AIC
Current License: CC BY-SA 3.0
5 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jan 29, 2022 at 20:55 | comment | added | Francis Laclé | To expand on the comment of @agcala including the bias, the computation for the number of parameters is $s_{j+1}\times{\left(s_{j}+1\right)}$ where $j$ is the index of a layer and $s$ the number of nodes in that layer. So, in this example, we have $\left(4\times{4}\right)+\left(2\times{5}\right)=26$. | |
| Jan 9, 2020 at 13:24 | comment | added | agcala | Total number of connections would be 26 if bias nodes were included. | |
| Jul 24, 2019 at 13:50 | comment | added | Funkwecker | Connections can be non-unique (see ieeexplore.ieee.org/document/714176). Hence, is it okay to simply count the connections? Maybe we should distinguish between parameter and hyperparameter? | |
| Sep 29, 2015 at 9:22 | vote | accept | Funkwecker | ||
| Oct 12, 2020 at 11:47 | |||||
| Sep 28, 2015 at 17:59 | history | answered | gung - Reinstate Monica | CC BY-SA 3.0 |